Insulated-gate bipolar transistors (IGBT) have been widely applied as power switching devices, motor PWM control inverters, etc. In general, the capacity of an IGBT increases with its size. Recently, in order to increase the capacity, a plurality of IGBTs are incorporated in a module structure within a package.
In a conventional package assembly for an IGBT-based semiconductor device, an emitter electrode as a main electrode and a gate electrode as a control electrode are placed side by side on a first principal surface of the device. A collector on a second principal surface is directly mounted on a metal base that also serves as a heat radiator. The emitter electrode and the gate electrode on the first principal surface must be connected to the exterior by separate external lead terminals. As a result, a conventional package assembly structure requires not only the metal base on the collector side, but also external lead terminals for the emitter and gate electrodes on the top part of the package. The connection of the emitter and gate electrodes to the external lead terminal is generally accomplished by wire bonding with an aluminum wire of approximately 300 .mu.m in diameter.
With the above-described conventional assembly structure, although heat dissipation is sufficient on the collector side, virtually no heat dissipation occurs at the emitter side. Such a lack of heat dissipation at the emitter side greatly reduces the current capacity of the IGBT. Moreover, an IGBT module having a large current capacitance normally includes a myriad of IGBT's, resulting in as many as several hundred bonding wires connected to the collector electrode. The large number of bonding wires cause a high level of internal wiring inductance, which may result in a large voltage surge during a switching operation of the IGBT module.
In an attempt to solve heat dissipation and internal inductance problems due to the above-described assembly structure, it is conceivable, as in the case of the conventional pressure contact type semiconductor device, to incorporate an IGBT in a flat package, where the collector and emitter electrodes that are formed on its respective principal surfaces are in planar contact with the respective top and bottom electrode plates of the package. However, such an IGBT is constructed such that the emitter electrode extends over an insulating layer that covers the gate electrode. When the electrode plates are brought in pressure contact with the emitter electrode, the plates also exert a pressure on the gate electrode, possibly resulting in destruction of the gate electrode structure. Thus, the above pressure contact type assembly structure is not practical.
Furthermore, where a pressure contact type flat package is used in a composite device in which an IGBT and a freewheel diode are incorporated in a package, additional problems arise. First, the wafer thicknesses of the IGBT and the freewheel diode, as required from electrical characteristics, are different from each other. When these different types of semiconductor elements of different thicknesses are incorporated side by side in a flat package, there occurs a difference in heights of the elements. That is, the electrode surfaces of the two semiconductor elements are uneven and form a step opposite the electrode plate of the package which is normally brought into planar contact with the respective semiconductor elements. Because of the unevenness of the surfaces of the elements, the electrode plate does not exert a uniform pressure on them. It is known from experiments that, if the difference between the heights of the semiconductor elements is not within .+-.50 .mu.m, the heat dissipation performance and the electrical characteristics of the semiconductor device are much deteriorated.
In addition, while conventional IGBTs generally use a wafer (called "epitaxial wafer") that is formed by sequentially growing n+ and n- layers on a p+ silicon substrate by epitaxy to attain desired switching characteristics, freewheel diodes use a wafer that is formed by growing an n+ layer by epitaxy on an n- silicon substrate that is normally produced by an FZ or MCZ method to reduce the cost. In the epitaxial wafer and the FZ or MCZ wafer, the thickness of the n- layer is set at an optimum value which produces the desired operating characteristics.
In the epitaxial wafer employed in the IGBT, a highly doped substrate is used as the p+ silicon substrate so that its small resistance has virtually no effect on the saturation voltage. The thickness of the p+ substrate is set to secure the required strength. On the other hand, in the diode using the FZ or MCZ wafer, the n+ epitaxial layer formed on the n- silicon substrate is thin and, therefore, the total thickness of the substrate is primarily determined by the thickness of the n- layer. As a result, the IGBT generally has a greater height than the freewheel diode, and their surfaces are uneven when placed side by side on a flat electrode plate for assembling.
For the foregoing reasons, there is a need for a pressure contact type semiconductor device, and an assembly method thereof, which incorporates a plurality of IGBTs in a flat package, and which provides improved heat dissipation, and a small internal wiring inductance. There is a further need for a composite device in which an IGBT and a freewheel diode are incorporated in a flat package, and which provides uniform pressure contact between the semiconductor elements and a common electrode plate of the package to avoid deterioration of their operating characteristics.